Tables

Descriptive tables

Global differences by:, age, sex, world region

IHME WHO Dif % difference (WHO / IHME) % difference (IHME / WHO)
Total HIV+TB only 211604 389042 177438.13321 83.85% -45.61%
TB only 1111312 1379440 268128.44955 24.13% -19.44%
Total TB 1322916 1768482 445566.58275 33.68% -25.19%
Adults HIV+TB only 177567 348026 170458.90473 96% -48.98%
TB only 1075691 1210620 134929.12946 12.54% -11.15%
Total TB 1253257 1558645 305388.03419 24.37% -19.59%
Children HIV+TB only 34037 41016 6979.22848 20.5% -17.02%
TB only 35621 168821 133199.32009 373.93% -78.9%
Total TB 69659 209837 140178.54857 201.24% -66.8%
Female HIV+TB only 78110 143496 65386.51804 83.71% -45.57%
TB only 367764 352488 15276.43876 -4.15% 4.33%
Total TB 445874 495984 50110.07929 11.24% -10.1%
Male HIV+TB only 99457 204471 105013.90757 105.59% -51.36%
TB only 707927 858132 150205.56821 21.22% -17.5%
Total TB 807383 1062603 255219.47578 31.61% -24.02%
AMR HIV+TB only 579 620 41.31917 7.14% -6.66%
TB only 2036 1914 122.37010 -6.01% 6.39%
Total TB 2615 2534 81.05093 -3.1% 3.2%
EMR HIV+TB only 165 533 368.30203 223.05% -69.04%
TB only 14658 14572 85.51575 -0.58% 0.59%
Total TB 14823 15106 282.78629 1.91% -1.87%
EUR HIV+TB only 212 374 161.68749 76.15% -43.23%
TB only 2383 2999 615.93832 25.85% -20.54%
Total TB 2595 3373 777.62581 29.97% -23.06%
SEA HIV+TB only 19310 28870 9560.04060 49.51% -33.11%
TB only 333250 345889 12639.43214 3.79% -3.65%
Total TB 352560 374759 22199.47275 6.3% -5.92%
WPR HIV+TB only 2057 2010 47.13348 -2.29% 2.35%
TB only 39055 28351 10704.20283 -27.41% 37.76%
Total TB 41112 30361 10751.33632 -26.15% 35.41%

Analytical tables

Table with model output for estimating likelihood or magnitude of difference in estimates by HIV, age, sex, and region.

This section is unfinished.

Graphs

Descriptive graphs

Standardized difference

Rankings of highest absolute and standardized differences for IHME and WHO.

Standardized difference for incidence

AB metric (a-b) / (a+b)

Rankings of highest absolute and standardized differences for IHME and WHO.

Maps

The below scatterplot shows the correlation between WHO (x-axis) estimates and IHME (y-axis) estimates, with each point colored by its (WHO-defined) region.

Analytical for adjusted_stand_dif

In the following four charts, Libya has been excluded as an outlier.

a) HIV prevalence among TB cases

b) Rifampicine resistance (MDR prevalence)

b.i) Rifampicine resistance (MDR prevalence using reported cases)

c) CDR (case detection rate per WHO)

c.i) CDR (case detection rate per IHME)

c.ii) Correlation between different case detections rates

d) CFR (case fatality rate)

Prevalence survey’s effect

Association of prevalence survey and standardized difference

# A tibble: 2 x 2
  prevsurvey median_adjusted_stand_dif
       <dbl>                     <dbl>
1       0                       - 5.20
2       1.00                     16.5 

Analytical for stand_dif_inc_adj

a) HIV prevalence among TB cases

b) Rifampicine resistance (MDR prevalence)

b.i) Rifampicine resistance (MDR prevalence using reported cases)

c) CDR (case detection rate per WHO)

Error: <text>:16:0: unexpected end of input
14: # Those countries with low case detection (oftentimes have prev survey),
15: # WHO estimates more deaths than IHME
   ^

c.i) CDR (case detection rate per IHME)

d) CFR (case fatality rate)

Prevalence survey’s effect

Association of prevalence survey and standardized difference

Analytical for adjusted_ab

a) HIV prevalence among TB cases

b) Rifampicine resistance (MDR prevalence)

b.i) Rifampicine resistance (MDR prevalence using reported cases)

c) CDR (case detection rate per WHO)

c.i) CDR (case detection rate per IHME)

d) CFR (case fatality rate)

Prevalence survey’s effect

Association of prevalence survey and standardized difference

Modeling

Linear regression to estimate effect of prevalence survey on absolute difference in cases (WHO minus IHME), adjusting for region.

95% confidence intervals

Linear regression to estimate effect of prevalence survey on adjusted standardized difference in cases, adjusting for region.

95% confidence intervals

Interactive map

Relative difference (function of maximum estimate)

(Unfinished)

Map of world

Map of Mozambique

Correlation coefficients

Correlation of adjusted stand diff with a) HIV prevalence, CDR by both, CFR, MDR prevalence.

cor(df$adjusted_stand_dif, df$newrel_hivpos, use = 'complete.obs')
[1] 0.09204849
cor(df$adjusted_stand_dif, df$gb_c_cdr, use = 'complete.obs')
[1] -0.3688292
cor(df$adjusted_stand_dif, df$cdr_ihme, use = 'complete.obs')
[1] 0.4355758
cor(df$adjusted_stand_dif, df$case_fatality_rate_2014, use = 'complete.obs')
[1] -0.1353054
cor(df$adjusted_stand_dif, df$case_fatality_rate_2012_to_2014, use = 'complete.obs')
[1] -0.1171681
cor(df$adjusted_stand_dif, df$case_fatality_rate_2015, use = 'complete.obs')
[1] -0.1586533
cor(df$adjusted_stand_dif, df$case_fatality_rate_2014_new, use = 'complete.obs')
[1] -0.1587409
cor(df$adjusted_stand_dif, df$case_fatality_rate_2012_to_2014_new, use = 'complete.obs')
[1] -0.1147299
cor(df$adjusted_stand_dif, df$case_fatality_rate_2015_new, use = 'complete.obs')
[1] -0.1760415
cor(df$adjusted_stand_dif, df$case_fatality_rate_2015_adjusted, use = 'complete.obs')
[1] -0.1825428
cor(df$adjusted_stand_dif, df$p_mdr_new, use = 'complete.obs')
[1] 0.05559062
cor(df$adjusted_stand_dif, df$reported_mdr, use = 'complete.obs')
[1] -0.0131539

Hypothesis testing on prevalence survey

Does region affect likelihood of having a prevalence survey?

xt <- table(df$prevsurvey, df$who_region)
xt
   
    AFR AMR EMR EUR SEA WPR
  0  37  37  20  52   8  22
  1  10   0   2   0   3   4
chisq.test(xt)

    Pearson's Chi-squared test

data:  xt
X-squared = 21.511, df = 5, p-value = 0.0006482

Does having a prev survey affect the adjusted stand diff?

t.test(x = df$adjusted_stand_dif[df$prevsurvey == 0],
       y = df$adjusted_stand_dif[df$prevsurvey == 1])

    Welch Two Sample t-test

data:  df$adjusted_stand_dif[df$prevsurvey == 0] and df$adjusted_stand_dif[df$prevsurvey == 1]
t = -2.1643, df = 21.066, p-value = 0.04207
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 -39.2150725  -0.7865973
sample estimates:
mean of x mean of y 
-4.545393 15.455442 

Supplemental charts

April 18, 2018

Alberto Method: top and bottom 10 stand diff (children only)

Martien Method: top and bottom 10 stand diff (children only)

Concordance between adult and child measures (Alberto method)

Concordance between adult and child measures (Martien method)

Alberto Method: top and bottom 10 stand diff (HIV only, all ages)

Martien Method: top and bottom 10 stand diff (HIV, all ages)

Effect of prevalence survey

Is CDR associated with stand diff in the linear regression among those countries without prevalence survey?

x <- df
x$gb_c_cdr[!is.finite(x$gb_c_cdr)] <- NA
x$original_stand_dif[!is.finite(x$original_stand_dif)] <- NA
fit <- lm(gb_c_cdr ~ original_stand_dif, data = x[x$prevsurvey == 0,])
summary(fit)

Call:
lm(formula = gb_c_cdr ~ original_stand_dif, data = x[x$prevsurvey == 
    0, ])

Residuals:
   Min     1Q Median     3Q    Max 
-42.50 -12.33   4.16  10.80  23.21 

Coefficients:
                   Estimate Std. Error t value             Pr(>|t|)    
(Intercept)         76.0320     1.1661  65.201 < 0.0000000000000002 ***
original_stand_dif  -0.3555     0.1123  -3.166              0.00187 ** 
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 14.35 on 151 degrees of freedom
  (23 observations deleted due to missingness)
Multiple R-squared:  0.06226,   Adjusted R-squared:  0.05605 
F-statistic: 10.03 on 1 and 151 DF,  p-value: 0.001868

Yes. Among countries without a prevalence survey, a 1 unit increase in the standardized difference is associated with a 0.355 unit decrease in the CDR. This is significant (p < 0.01).

More visualizations

Afternoon of Wednesday, April 19. All charts are WHO minus IHME

a) absch: w_both_014_tbtotal_nd - i_both_014_tbtotal_nd

Country Value
India 49508.8443
Nigeria 32003.9550
Indonesia 12752.4252
Bangladesh 8616.7562
Tanzania 5992.6238
Dem. Rep. Congo 5147.3671
China 3214.4091
Myanmar 3169.4789
Pakistan 2334.5255
Angola 2089.3906
Zimbabwe -2191.6025
South Africa -1524.9994
Malawi -1022.3595
Uganda -504.5276
Burkina Faso -499.9316
Lesotho -415.1728
Rwanda -378.9136
Burundi -279.2630
Swaziland -223.2993
Namibia -170.1219

standardized

Country Value
North Korea 100.00000
Bangladesh 99.41480
Timor-Leste 99.01265
Papua New Guinea 98.80690
Libya 97.09310
Viet Nam 97.08583
Thailand 95.00197
United Arab Emirates 94.13214
Haiti 93.15330
Myanmar 92.41885
Rwanda -68.52748
Zimbabwe -55.26266
Azerbaijan -53.23740
Burkina Faso -49.28143
Japan -47.64091
Swaziland -40.43862
Honduras -36.12995
Namibia -33.76727
Malawi -30.59634
Eritrea -30.44260

Downloadable table with full data

b) absad: (w_f_15plus_htb_nd + w_m_15plus_htb_nd + w_f_15plus_tb_nd + w_m_15plus_tb_nd) - ( i_both_15plus_tb_nd + i_both_15plus_htb_nd)

Country Value
Nigeria 184617.353
Bangladesh 41246.468
Tanzania 32279.694
South Africa 30633.208
Mozambique 27458.985
Dem. Rep. Congo 20863.048
Indonesia 13369.005
North Korea 11553.577
Ghana 8202.510
Madagascar 7978.496
India -28812.449
Ethiopia -22502.583
China -16752.815
Philippines -9655.750
Zimbabwe -8890.743
Nepal -5859.519
Viet Nam -4999.429
Uganda -4576.158
Burkina Faso -4337.049
Niger -3617.583

standardized

Country Value
Nigeria 97.94014
Marshall Islands 88.17223
Timor-Leste 87.20306
Papua New Guinea 85.49242
Libya 85.13396
North Korea 84.44479
Greenland 77.69338
Iceland 76.08632
Sudan 66.45726
Congo 65.73305
Azerbaijan -100.00000
Egypt -86.46321
Macedonia -83.75589
Saint Lucia -80.58197
Rwanda -78.99093
Syrian Arab Republic -71.15123
Burkina Faso -67.09132
Comoros -67.08868
Eritrea -66.89072
Honduras -65.43858

Downloadable table with full data

c) abshiv: w_both_all_htb_nd - i_both_all_htb_nd

Country Value
Nigeria 52805.3129
South Africa 29593.6434
Indonesia 19480.6093
Tanzania 18779.4125
Mozambique 17870.0413
India 9962.0335
Dem. Rep. Congo 9497.6106
Zambia 7498.0137
Angola 5670.8262
Ghana 3740.3204
Zimbabwe -6432.0229
Ethiopia -4338.3535
Uganda -2717.5779
Botswana -1463.5821
Namibia -944.1877
Viet Nam -920.1390
Rwanda -748.9198
Swaziland -744.3264
Burkina Faso -738.9729
Burundi -640.2910

standardized

Country Value
Madagascar 96.74147
Bhutan 93.36600
Afghanistan 93.35732
Moldova Republic of 88.93088
Nigeria 85.30343
Cape Verde 82.80465
Papua New Guinea 78.06040
Uzbekistan 77.99681
Congo 76.74764
Mauritius 73.54996
Turkmenistan -100.00000
Chile -82.69580
Argentina -73.74461
Puerto Rico -68.08163
Japan -66.53380
North Korea -58.28064
Serbia -57.62913
Rwanda -57.54725
Burkina Faso -57.53504
Iran Islamic Republic of -55.40047

Downloadable table with full data

d) abstb: w_both_all_tb_nd - i_both_all_tb_nd

Country Value
Nigeria 163815.995
Bangladesh 49751.746
Tanzania 19492.905
Dem. Rep. Congo 16512.804
North Korea 13321.155
Mozambique 11039.108
India 10734.362
Madagascar 8024.573
Afghanistan 7421.693
Indonesia 6640.821
Ethiopia -18311.726
China -13512.763
Philippines -9788.048
Nepal -5666.740
Zimbabwe -4650.323
Burkina Faso -4098.008
Niger -3269.260
Senegal -2887.683
Uganda -2363.108
Viet Nam -2331.514

standardized

Country Value
Nigeria 93.09254
Marshall Islands 86.33247
Timor-Leste 86.30530
Libya 86.10756
North Korea 84.53367
Papua New Guinea 84.33227
Greenland 74.17450
Gabon 73.73470
Iceland 73.35414
Sudan 66.26216
Azerbaijan -100.00000
Macedonia -79.75107
Rwanda -79.58858
Honduras -76.10209
Bahamas -72.93482
Saint Lucia -72.87303
Egypt -67.49221
Zimbabwe -64.95982
Burkina Faso -62.26468
Kuwait -61.81510

Downloadable table with full data

e) absall: w_both_all_tbtotal_nd - i_both_all_tbtotal_nd

Country Value
Nigeria 216621.308
Bangladesh 49863.225
Tanzania 38272.318
South Africa 29108.209
Mozambique 28909.149
Indonesia 26121.431
Dem. Rep. Congo 26010.415
India 20696.396
North Korea 13218.478
Angola 9910.123
Ethiopia -22650.079
China -13538.406
Zimbabwe -11082.346
Philippines -9435.607
Nepal -5476.979
Uganda -5080.686
Burkina Faso -4836.981
Niger -3757.964
Viet Nam -3251.653
Senegal -3147.204

standardized

Country Value
Nigeria 98.95851
Marshall Islands 91.15784
Timor-Leste 90.95049
Papua New Guinea 89.67219
Libya 89.20694
North Korea 87.77826
Greenland 77.96434
Iceland 76.32096
Sudan 69.42114
Laos 68.21180
Azerbaijan -100.00000
Macedonia -84.39432
Rwanda -79.30289
Saint Lucia -72.67408
Egypt -68.95801
Burkina Faso -66.12854
Honduras -64.72053
Eritrea -64.18513
Kuwait -63.76456
Comoros -61.38086

Downloadable table with full data

Big table for appendix

Missing data in CDR, CFR, HIV

# CDR
df$country[is.na(df$cdr_ihme)]
[1] "Antigua and Barbuda" "Turkmenistan"        "Qatar"              
[4] "Bahrain"             "Comoros"             "Virgin Islands U.S."
# CFR
df$country[is.na(df$case_fatality_rate_2015)]
 [1] "Sudan"                            "Cambodia"                        
 [3] "Malta"                            "Antigua and Barbuda"             
 [5] "Gambia"                           "Djibouti"                        
 [7] "Turkmenistan"                     "Qatar"                           
 [9] "Luxembourg"                       "Bosnia and Herzegovina"          
[11] "Canada"                           "Grenada"                         
[13] "Japan"                            "Italy"                           
[15] "Bahrain"                          "American Samoa"                  
[17] "Switzerland"                      "Ethiopia"                        
[19] "Belize"                           "Greece"                          
[21] "France"                           "Saint Vincent and the Grenadines"
[23] "Comoros"                          "Saint Lucia"                     
[25] "Bermuda"                          "Virgin Islands U.S."             
df$country[is.na(df$case_fatality_rate_2014)]
 [1] "Sudan"                            "Cambodia"                        
 [3] "Malta"                            "Antigua and Barbuda"             
 [5] "Gambia"                           "Djibouti"                        
 [7] "Turkmenistan"                     "Qatar"                           
 [9] "Luxembourg"                       "Bosnia and Herzegovina"          
[11] "Canada"                           "Grenada"                         
[13] "Japan"                            "Italy"                           
[15] "Bahrain"                          "American Samoa"                  
[17] "Switzerland"                      "Ethiopia"                        
[19] "Belize"                           "Greece"                          
[21] "France"                           "Saint Vincent and the Grenadines"
[23] "Comoros"                          "Saint Lucia"                     
[25] "Bermuda"                          "Virgin Islands U.S."             
# HIV
df$country[is.na(df$newrel_hivpos)]
 [1] "Iceland"                "Algeria"               
 [3] "Mauritania"             "Austria"               
 [5] "Turkmenistan"           "Qatar"                 
 [7] "Poland"                 "Hungary"               
 [9] "Luxembourg"             "United Kingdom"        
[11] "Bosnia and Herzegovina" "Korea Republic of"     
[13] "Germany"                "Bulgaria"              
[15] "Italy"                  "Bahrain"               
[17] "American Samoa"         "Spain"                 
[19] "Denmark"                "Switzerland"           
[21] "Greece"                 "Croatia"               
[23] "France"                 "Sweden"                
[25] "Cyprus"                 "Comoros"               
[27] "Virgin Islands U.S."   

More correlations

Using stand_dif ( (a-b)/(a+b)/2 )

Standardized difference with CDR WHO

Standardized difference with CDR GBD

Standardized difference with CFR

(using variable case_fatality_rate_2015_adjusted)

Comparison of different case fatality rate measures

Alberto asked: can you look at the code and compare case_fatality_rate_2015_adjusted and case_fatality_rate_2015

They are identical except that the adjusted rate has 17 missings, whereas the non adjusted rate has 26 missings.

Standardized difference with CFR

(using variable gb_cfr)

Standardized difference with Reported HIV prevalence

Standardized difference with MDR prevalence

(using reported_mdr)

Standardized difference with MDR prevalence

(using p_mdr_new)

Request A: Correlations with original_stand_diff

CDR WHO

stand_dif_compare(x = df$original_stand_dif, 
                  y = df$gb_c_cdr,
                  yl = 'CDR WHO')

CDR GBD

stand_dif_compare(x = df$original_stand_dif, 
                  y = df$cdr_ihme,
                  yl = 'CDR GBD')

HIV prevalence

stand_dif_compare(x = df$original_stand_dif, 
                  y = df$p_hiv_of,
                  yl = 'HIV prevalence')

MDR prevalence

stand_dif_compare(x = df$original_stand_dif, 
                  y = df$p_mdr_new,
                  yl = 'MDR prevalence')

Request B: Box plot of association of prevalence survey with stand diff (not original_stand_dif)

Difference using t-test (for p-value)


    Welch Two Sample t-test

data:  df$stand_dif[df$prevsurvey == 0] and df$stand_dif[df$prevsurvey == 1]
t = -2.1643, df = 21.066, p-value = 0.04207
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 -0.78430145 -0.01573195
sample estimates:
  mean of x   mean of y 
-0.09090786  0.30910884 

prev_survey and original_stand_dif

Difference using t-test (for p-value)


    Welch Two Sample t-test

data:  x$original_stand_dif[x$prevsurvey == 0] and x$original_stand_dif[x$prevsurvey == 1]
t = -1.1965, df = 22.444, p-value = 0.244
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 -8.447386  2.261924
sample estimates:
mean of x mean of y 
 1.027875  4.120606